[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]


To: dns op wg <dnsop@cafax.se>, Ted Hardie <Ted.Hardie@nominum.com>
From: Alain Durand <alain.durand@sun.com>
Date: Fri, 21 Dec 2001 18:50:30 -0800
In-Reply-To: <E16H5eN-000FtJ-00@rip.psg.com>
Sender: owner-dnsop@cafax.se
Subject: Re: another iesg comment on draft-ietf-dnsop-hardie-shared-root-server-06.txt

At 08:02 AM 12/20/2001 -0800, Randy Bush wrote:
>The document doesn't seem to say anything about path MTU
>discovery (and potentially other ICMP error issues) for
>packets sent with an anycast source address.
>
>Since an ICMP "packet too big" will be sent back by a router to the anycast
>address it might not be delivered to the anycast member that actually
>sent the packet which might cause retransmissions with persistent
>packet loss.
>
>This fix is simply to point out that the responses from an anycast
>address must not use path MTU discovery i.e. must not have the DF bit set.
>
>I don't know of other ICMP errors that have similar issues.

I know this draft specificaly applies to IPv4 and not IPv6, but
in v6, the DF bit is gone, routers are not doing fragmentation
and pMTU discovery is always on....

If one day we could get IPv6 to relax the rules on anycast,
It would be nice to be able to re-use this document without too
many changes.

         - Alain.


Home | Date list | Subject list